Could ChatGPT Generate Lucrative Results in Online Search and Product Discovery?
MACH Alliance experts unpack the potential business benefits of generative AI for retail and e-commerce.
Less than six months after the official public release of ChatGPT, retailers are already taking generative AI technology into their fitting rooms to try it on for size.
Among the flurry of generative AI news in the sector:
- French grocer Carrefour released a demo video of a ChatGPT avatar it created to answer shoppers’ questions
- Amazon disclosed it’s been developing software similar to ChatGPT for a few years
- Chinese retailer JD.com plans to launch its own generative AI tool
- Shopify has integrated ChatGPT into the shopping assistant feature of its Shop app
With so many potential applications for this new tech capability, narrowing the focus is essential. For example: What impact could generative AI have, in particular, on online search and product discovery? Three experts recently weighed in on that and the overall future of retail search during a virtual roundtable hosted by the MACH Alliance.
Changing the game
Compared to the AI currently used in online search, generative AI is far less text-based.
Besides generating results from keywords, it also produces search results based on input images, video, and audio. It can even ask a shopper questions to refine their search in a much more human, conversational way than the chatbots we’ve become used to.
Panelist Eli Finkelshteyn, CEO of Constructor said these expanded capabilities will liberate search and product discovery from the limited UI of typing in keywords.
“Now all of a sudden, you’re able to start understanding these longer-form questions. That’s getting us into a new era of product discovery I really don’t think existed beforehand, because people weren’t able to search in that way,” said Finkelshteyn.
“It’s not just about letting people search in a new way,” he added. “It’s actually about generating revenue that wasn’t generated before because people couldn’t find those things [through traditional search technology].”
Besides widening the breadth of search and discovery, generative AI might also improve CX. That’s because the interactive, conversational interface it offers is more natural and engaging while requiring less effort from consumers, said roundtable guest Nilay Oza, CEO of Klevu.
“Google primarily was, and is still, a one-query engine: you write a query, you get an answer. With the likes of ChatGPT, you are actually becoming a multi-query engine. So you can actually have a system where it’s asking you a question to further help it understand,” Oza said.
Better B2B search
Most AI can personalize B2C search based on an individual’s data. Personalizing a B2B search is tougher, however, because an individual’s data might not be relevant to their business.
According to the third panelist, Raj De Datta, CEO of Bloomreach, generative AI could tackle this problem. He explained that generative AI needs very little contextual data to produce relevant search results, which makes it particularly promising for B2B search.
“[Generative AI] is able to draw relationships between things that previously would not be naturally obvious. That means we don’t necessarily have to be reliant on depth of data per query or per user to return results. It unlocks a cohort of searches for whom it’s been hard, with the data sets we’ve had previously, to return great results,” he said.
The next frontier
Once generative AI masters conversational engagement with consumers across multiple touchpoints in various content forms, Oza believes it will eventually enable opinion-based search and product discovery. He envisions it serving up data-informed, contextually relevant, expert advice to consumers.
“We are now entering into a dialogue phase [of search]. The next stage after that would be more of an opinion-based discovery where the technology or the software is actually giving its opinion to you based on certain logic,” Oza predicted.
Finkelshteyn suggested generative AI could transform the way search results are visually displayed on-screen.
Here’s how: generative AI allows retailers to create incredibly life-like product images and videos based solely on data input – without taking a single photo, hiring any graphic designers, or shooting any footage. With this capability at their disposal, Finkelshteyn said online search results could potentially mimic the visual merchandising strategy used in stores.
“When I look at the presentation of search results right now, we’re still looking at these columns and grids of products. I think this is one of the places where e-commerce can really learn a lot from physical stores. They think very carefully about the clothes that are on the mannequins, how clothes are hung up, stuff like that. The feel you get for a brand when you see the different ways these products are displayed in-store, I think that’s something that hasn’t really been emulated online yet,” Finkelshteyn said.
Even more possibilities opened up for retailers on March 14, when the newest version of ChatGPT was publicly released. Compared to earlier iterations, ChatGPT-4 can deal with almost 33,000 tokens of information at a time versus only 4,100 tokens. It’s also multilingual, able to accurately answer questions in 26 different languages.
With retailers now entering the earliest phase of generative AI adoption, they’ve still got a lot of exploration ahead of them to tailor it to the needs of their businesses and customers.
Christine Wong
Senior Technology Staff Writer, Orium
I've been telling enterprise technology stories for almost three decades in print, online, and on television. I started out in journalism, covering the telecom boom, the birth of social media and the emergence of digital commerce. I'm always looking for the human angle in every technology story I write.